Fragmented Packets and packet size

Andy M5ZAP

While trying to fault find an issue with the waterfall and spectrum not displaying on a remote smartlink connection.....

I observed while monitoring the local network traffic on my LAN that the Flex was sending ipv4 packets of 1514 bytes. As my PC has an MTU of 1500 bytes fragmentation and reassembly has to occur.

My router also has a fixed MTU of 1500 bytes on the outbound interface so all packets sent over the internet are being fragmented.

When trying to connect remotely, wireshark shows reassembly errors. "netstat-s" shows a huge number of packets requiring reassembly and zero successfully reassembled.

I understand from conversation with a core network engineer at an ISP that any packet size greater than 1500 bytes is not best practice and can result in issues.

Is this normal to have an MTU on the flex packets greater than 1500 and can it be reduced.

Tim - W4TME

Packets sent from the radio are not greater than the max for an Ethernet frame (1518 bytes).  Devices will list an MTU of 1500 when it really is 1518 (14 bytes of Ethernet header and 4 bytes of CRC, leaving 1500 bytes of payload).

The VITA-49 data sent for display data is greater than the 1500 byte payload of an Ethernet packet so the TCP/IP stack on the radio is fragmenting the data and the TCP/IP stack on the client is reassembling it.  This is a standard and customary way of handling large data streams over a packet network.

If you are getting the errors you are reporting, the problem is not with the radio, but most likely with your router not allow fragmented packets to pass.  This can be a global setting or implemented as a firewall rule.  Since the maximum sized fragmented packets are not allowed to pass and the ones that are smaller do, Wireshark reports this as reassembly errors.

We have had at least one customer that has this issue and it was resolved when you configured their firewall/router to allow fragmented packets to pass.

Andy M5ZAP

Hi Tim,

Thanks for the quick reply, I will try another router. The current one is a BT Homehub which is a very popular router in the UK so I would of expected more reported issues.

I can successfully connect via my iOS device and 4G, if my router was blocking fragmented packets wouldn't  that connection also be affected ?

Tim - W4TME

It is usually the inbound router (on the client end) that will not allow the packets in.  Most firewalls do not filter on outbound connections.

Andy M5ZAP

Ok thanks, I had assumed incorrectly you were referring to the outbound router.

Tim - W4TME

No problem.  I may not have been completely clear in my previous post.

Andy M5ZAP

Might be a **** question but why cant the packets be made smaller than 1500 so that fragmentation doesn't occur.

Also don't the packets being fragmented cause problems with latency if they take a different route etc.

Tim - W4TME

I'll answer the last question first.  No.  

The route the packets take is really insignificant in terms of variable latency.  Internet routing (BGPv4) is very efficient and will usually send all packets to their destination using the lowest latency path (assuming that the underlying IGP is a distance-vector routing protocol).  The other part of your question really deals with data transfer efficiency, so I'll address that next.

Using smaller packets results in inefficient data transfer due to the additional packet overhead and the additional CPU required to manually segment the data payload, something that the OS kernel does much more efficiently in the TCP/IP stack.

Andy M5ZAP

So the inefficiency of making the packet 18 bytes smaller is greater than the overhead of having to fragment and reconstruct the packets all the time when they leave the flex ?

Varistor

Try MTU 1492. This is not a random number.

Tim - W4TME

This isn't an MTU issue.

Andy M5ZAP

So the inefficiency of making the packet 18 bytes smaller is greater than the overhead of having to fragment and reconstruct the packets all the time when they leave the flex ?

Tim - W4TME

Yes.

Eric-KE5DTO

Note that the size of the display packets is directly related to the width of the display.  If you're right on the edge of a boundary, you can make the display more narrow to test that boundary.

Andy M5ZAP

Thanks Eric I will test that and report back. Any guide as to what width should result in a vita packet size that won't be fragmented ( I assume < 1500)
For my own interest how big can the vita packets be before the necessary fragmentation?

Tim - W4TME

For my own interest how big can the vita packets be before the necessary fragmentation? 

When the data payload is > 1500 bytes

KB4AAA

I have spectrum internet as far as I can tell they are the cause of no display of waterfall and spectrum display while using remote smartlink. I have kinda of a workaround to at least display the spectrum so it will be usable without using VPN. If you put Smartsdr in a window and shrink the width down (height dosen't seem to matter)  and open the radio dock on the right side. The spectrum display will start working.
It even works with all 4 slices open

You can even undock a slice and it will work unless you make it to wide

Hopefully this may help others to work remotely until there is some kinda fix for the fragmentation.